As the risk for pancreatic ductal adenocarcinoma (PDAC) has increased to a higher rate in veterans relative to that in the general population, there is an urgent need to develop effective therapies to treat PDAC for veterans in the VA system. Recently, cancer immunotherapy has shown great promise in several cancers, but not in PDAC. Part of the reason for this is the heterogeneity of PDAC and lack of tailoring of immunotherapy to individual tumor subtypes. PDAC patient stratification for therapy remains in its infancy, and a reliable method to deconvolute complex tumor composition to stratify PDAC subtypes has not been recognized. Recently, we participated in whole genome sequencing of PDAC specimens, which provides the basis for classifying PDAC into four subtypes based upon patterns of genomic structural variation (Nature, 2016). Among these, the immunogenic subtype accounts for 30% of 178 PDAC samples in the TCGA database and is characterized by upregulated immune cell networks, which could indicate differential responses to immunotherapy. In addition, the Epigenomic Deconvolution (EDec) method, first developed by our co-investigator Dr. Milosavljevic?s group, provides valuable information about cell type composition of tumors and cell-type specific gene expression (Cell Rep. 2016). When applied to immunogenic subtype PDAC tumors, EDec reveals an immunosuppressive microenvironment characterized by the highest Foxp3 expression among all four subtypes. Cancer cells falling into the immunogenic subtype also show the highest mesothelin (MSLN) expression. Therefore, we hypothesize that PDACs with an immunogenic profile could be a target subgroup that is responsive to immunotherapy either by MSLN virus-like particle (VLP) vaccination or combination therapy of VLPs and an immune checkpoint inhibitor. We will test our hypothesis in pre-clinical animal models that best recapitulate human PDAC patient?s response to immunotherapy, including patient-derived tumor xenografts (PDX) and humanized mouse models. Our preliminary data have shown that our anti-MSLN VLP vaccine is effective against MSLN-high expressing PDX in a humanized NSG mouse model (PDX-hu-NSG). Based on our strong preliminary results, we propose to develop an effective cancer immunotherapeutic approach by combining three highly synergistic innovations: (1) A novel epigenetic deconvolution method to stratify PDAC tumors; (2) PDX-hu-NSG models; and (3) Combination therapy of MSLN-VLP vaccine plus anti-PD-1 antibody. We propose two specific aims. In Aim 1, we will determine whether the immunogenic subtype of PDAC is responsive to MSLN-VLP vaccine in PDX-hu-NSG model. Here, we will use EDec method to stratify VA PDACs by subtype and then determine MSLN-VLP vaccine efficacy in specific PDAC subgroups in humanized PDX mouse model. In Aim 2, we will determine whether combination therapy with anti-PD-1 Ab enhances MSLN VLP vaccine responses and efficacy in onco-humice model. We will also determine specific tumor infiltrating subset of cells that are responsible for the effective combination therapy. Furthermore, potential side-effect of the combination therapy will also be evaluated. Our findings will provide preclinical evaluation of the therapeutic efficacy of an innovative precision immunotherapy for PDAC in humanized mice without putting patients at risk. The project will provide understanding of molecular, cellular, and tissue-level responses to therapy, a key step towards improved outcomes in PDAC through patient stratification for therapy.